Marcus Gartner
Marcus Gartner
A classic 5 year ["temporary fix"](https://github.com/cockroachdb/cockroach/pull/30216#issuecomment-422041328) finally addressed. cc @RaduBerinde.
@DrewKimball I added a new commit to address one assumption in the join reader that broke because of this. PTAL.
Hmm I wonder if this could cause a regression in the row-based engine in the case that a filter like `i IN (1 , 3, …)` is pushed to another...
I re-tested this and it did cause a significant regression when evaluating a distributed filter like in `SELECT * FROM t WHERE a IN (1, 2, 3, 4, ..., 9999);`....
~~The first two commits (currently) are from #119557.~~ The flow setup benchmarks show some improvements: ``` name old time/op new time/op delta FlowSetup/vectorize=true/distribute=true-10 51.5µs ± 0% 51.3µs ± 1% ~...
TFTR! bors r+
*[`pkg/util/timeutil/timer.go` line 110 at r1](https://reviewable.io/reviews/cockroachdb/cockroach/119595#-NrO2bGG90giCZ5iac5m:-NrfzaUF6XHWN54Go9e9:b-k4n5vg) ([raw file](https://github.com/cockroachdb/cockroach/blob/1a20c30b46d878db20dd7465da31eaa5a4adf602/pkg/util/timeutil/timer.go#L110)):* Previously, yuzefovich (Yahor Yuzefovich) wrote… On a quick search there are like 20 places where the timers are allocated on the "stack" and...
I guess we can't clear the whole buffer, but we can remember the previous size and `buf.Truncate`, right?
>Probably worth backporting this too, right? Yes - added the labels.